[16.0][FIX] l10n_nl_xaf_auditfile_export: parse EntityRef error

[astirpe]

When an account has an "&" character in its name, the XAF auditfile export fails.

Here are 2 different errors I was able to reproduce:

• b'xmlParseEntityRef: no name' (probably the one described in line 1653: b'xmlParseEntityRef: no name' #397)
• b"EntityRef: expecting ';'"

Steps to reproduce:

• install any chart of account, in this example we use l10n_nl
• create and confirm an invoice
• create an Audifile export and verify that the xaf file is successfully generated
• modify the name of account 110000 Debiteuren (that for sure is listed in the auditfile), for example name it: "Debiteuren & test"
• create a new Audifile export and verify that an error is raised this time

After some investigations, it seems that value_to_html() returns a markup instead of a string. This seems making the etree.XMLParser() unhappy in case of any "&" character present. Converting the markup to string solves the issue. Not sure if there's a better way to solve it.

[thomaspaulb]

What do you mean by 'returns a markup' ?

If I look at what I think is the super function of the one you're modifying (I'm not 100% sure because there are several in ir_qweb_fields.py), and I naively run the code in Odoo shell, then I get this:

>>> from odoo.tools import pycompat, html_escape
>>> html_escape(pycompat.to_text('a & b'))
'a & b'
>>> str(html_escape(pycompat.to_text('a & b')))
'a & b'

I don't see a difference between before and after str() - both results are strings.

[astirpe]

@thomaspaulb did you try with the steps to reproduce?

[thomaspaulb]

@astirpe Not yet, but I believe you that I can reproduce it functionally that way. Before I go into the trouble of replicating the setup and setting a breakpoint to get the answer to my question, I'd ask you first. So, what do you mean by "returns a markup" ?

[astirpe]

@thomaspaulb it means that returns a type "Markup" instead of "String".

[thomaspaulb]

I did some tests with the help of #399, and indeed I'm running into the xmlParseEntityRef: no name error in cases where a field contains an ampersand.

However your fix only addresses cases where the widget is used. I found that changing xml.unescape() to str(xml) over here addresses the issue for all fields, regardless of the widget used (mine was here)

[astirpe]

I like your approach, I think is the correct way. Is it ok for you to fix it directly in your PR, so that we can close this one?

[thomaspaulb]

Let's keep both things separated.

[thomaspaulb]

Now done in #410

[ThijsvOers]

Tested the change and it works now

[hbrunn]

I'll merge this as it already fixes part of the problem

/ocabot merge

[OCA-git-bot]

Hi @hbrunn. Your command failed:

Required option bumpversion_mode for command merge. Possible values : major, minor, patch, nobump.

Ocabot commands

• ocabot merge major|minor|patch|nobump
• ocabot rebase
• ocabot migration {MODULE_NAME}

More information

• ocabot documentation
• OCA guidelines, specially the "Version Numbers" section.

[hbrunn]

/ocabot merge patch

[OCA-git-bot]

What a great day to merge this nice PR. Let's do it!
Prepared branch 16.0-ocabot-merge-pr-398-by-hbrunn-bump-patch, awaiting test results.

[OCA-git-bot]

Congratulations, your PR was merged at c356ef4. Thanks a lot for contributing to OCA. ❤️